Fig 9 - uploaded by Truc Thi Kim Nguyen
Content may be subject to copyright.
Source publication
We propose a novel multistage smoke detection algorithm based on
inherent optical characteristics such as diffusion, color, and texture
of smoke. Moving regions in a video frame are detected by an approximate
median background subtraction method using the diffusion behavior of
smoke. These moving regions are segmented by a fuzzy C-means (FCM)
clust...
Similar publications
Temporal action proposal (TAP) aims to detect the action instances’ starting and ending times in untrimmed videos, which is fundamental and critical for large-scale video analysis and human action understanding. The main challenge of the temporal action proposal lies in modeling representative temporal relations in long untrimmed videos. Existing s...
First-person action recognition is a challenging task in video understanding. Because of strong ego-motion and a limited field of view, many backgrounds or noisy frames in a first-person video can distract an action recognition model during its learning process. To encode more discriminative features, the model needs to have the ability to focus on...
We present an approach to learn voice-face representations from the talking face videos, without any identity labels. Previous works employ cross-modal instance discrimination tasks to establish the correlation of voice and face. These methods neglect the semantic content of different videos, introducing false-negative pairs as training noise. Furt...
Finger vein recognition is a promising personal identification technology due to its advantages over other existing biometric recognition. In this paper, a novel finger vein recognition method is proposed based on Log Gabor filter and Local Derivative Pattern. Through combining Log Gabor convolution magnitude and Local Derivative Pattern operator,...
Vehicle re-identification has become an important research topic for its application prospect in real-world, such as intelligent security system and intelligent traffic management. The current vehicle re-identification algorithms mainly run on image-based datasets, while video-based datasets are very rare in the community. Therefore we collect a da...
Citations
... In the training process of DL, the loss value of the current model is obtained by calculating the difference between the real label in the image and the output result of the neural network, and the loss value reflects the difference between the output result and the real label. Usually, the loss of target detection network consists of three parts, namely, category loss, detection frame loss and confidence loss (Nguyen & Kim, 2013). This target detection task is single target detection without category loss, so the loss of this detection task is composed of head frame and confidence loss. ...
Aiming at the problem that dangerous operation behaviors in the laboratory is difficult to identify by monitoring the video. An algorithm of dangerous operation behavior detection in multi-task laboratory based on improved YOLOv5 structure is proposed. Firstly, the algorithm enhances, adaptively scales, and adaptively anchors box computing on the input of YOLO network. Then convolution operation is carried out to strengthen the ability of network feature fusion. Finally, the GIoU_Loss function is used at the output to optimize the network parameters and accelerate the convergence of the model. The experimental results show that the algorithm performs well in real-time head localization, head segmentation, and population regression, with significant innovation and superiority. Compared with traditional methods, this algorithm has better accuracy and real-time performance and can more effectively achieve human operation behaviors detection in laboratory application environments.
... Jia et al. [11] proposed a smoke region segmentation method based on saliency detection. Calderara et al. [19] use image energy and color information to develop smoke detection system. ...
The spatial characteristics, movement characteristics and color characteristics of smoke are important features that distinguish them to other objects. In order to make full use of these three features, we proposed a deep convolutional network called Full High Resolution Network(FHRNet).This network consists of two parts: Spatio-Temporal-aware Sub-network (STS) and Color-aware Sub-network (CS). We build high -resolution residual symmetrical units and embed the two sub-networks to ensure the integrity of two dimensional features.In the STS, the residual symmetrical unit extracts the spatial semantic characteristics of smoke from every frame, and combine them into a feature sequence, then the spatio-temporal perceptron is used to extract the spatio-temporal characteristics of smoke to further improve the characteristic expression. In the CS, the color feature of picture is converted into color feature matrix, which is easier to make the residual symmetrical unit to extract the color feature of smoke. We constructed a smoke vedio datasets which have a diverse background to avoid producing over-fitting situation.The experimental results show that mthod we proposed can effectively extract the color features and the spatio-temporal features of smoke and our method can effectively detect smoke.
... Calderara et al. [20] use image energy and color information to develop smoke detection system. ...
Motion feature is the basic feature of smoke in the sequence. The spatial feature and color feature of smoke can be helpful to extract the motion feature of smoke. In this paper, we study how to express the three features of smoke and how to integrate these features effectively. We propose a Multi-Scale Cross Attention Net( MSCANet ) to extract the three features of smoke and establish the relationship between the three features. It is composed of Multi-Scale Convolution Module( MSCM ) and Cross Attention Module( CAM ). MSCM uses convolution kernels of different size to extract feature mapping in different receptive fields,which can capture the global and local features of each smoke image. We establish three MSCM to extract three features independently, and the convolution parameters of different MSCM are not shared, which can realize the independence of different feature extraction. CAM uses the attention mechanism to strengthen the expression ability of smoke features, and uses attention map multiply by two external features to realize the fusion and mutual guidance of the three features. We embed CAM into MSCM to realize the independence and correlation of three feature extraction. We conducted comparative experiments and ablation experiments in our own large dataset, and the accuracy can reach 84.94%. Experimental results show that the proposed method can effectively fuse the three features of smoke, so as to effectively detect smoke.
... The authors in [38] used Kalman filtering techniques in order to detect moving regions in cases of forest fire smoke detection. [28], [33], [41] . A review article with recent developments in video based fire detection has analysed recently used candidate region selection techniques for both smoke and fire [22]. ...
The burning issue is a very serious issue nowadays in our garments and industries sector.
The workers are facing the problem and losing valuable life. On the other hand, investors
are losing their hope in this sector. In this paper, we have propounded a vision-based
system which is capable to detect fire. We have developed a pipeline model consisting
of Background Subtraction, Color Segmentation, Special Wavelet Analysis & a Support
Vector Machine which will detect real-time fire and smoke. For the SVM model we have
trained the dataset in two ways. One is the different kind of fire image and other is the
image that looks like fire but it’s not fire. If the situation is breaking out of fire then the
system will immediately raise an alarm and an automatic SMS and email will be sent to
the authority and nearby fire station. In this study, the proposed strategy works on a very
large dataset of fire videos that have been collected both in real-life situations and from
the internet. In this SVM pipeline model shows the maximum accuracy is 93.33%. The
system can fulfill the precision and detect faster real-time fire detection. Its industrial
application will aid in the early detection of fires, as well as emergency management, and
so greatly contribute to loss prevention.
... Several smoke detection methods based on low-level features i.e., color, shape, and motion have been reported in the last decade. For instance, the authors in (Nguyen & Kim, 2013;Tung & Kim, 2011;Yuan, 2008) extracted motion features using optical flow, color, and image energy to identify smoke patterns inside an image. The classifiers used in these methods included back-propagation neural network and fuzzy C-mean for the clustering of smoke patterns. ...
Fire disaster throughout the globe causes social, environmental, and economical damage, making its early detection and instant reporting essential for saving human lives and properties. Smoke detection plays a key role in early fire detection but majority of the existing methods are limited to either indoor or outdoor surveillance environments, with poor performance for hazy scenarios. In this paper, we present a Convolutional Neural Network (CNN)-based smoke detection and segmentation framework for both clear and hazy environments. Unlike existing methods, we employ an efficient CNN architecture, termed EfficientNet, for smoke detection with better accuracy. We also segment the smoke regions using DeepLabv3+, which is supported by effective encoders and decoders along with a pixel-wise classifier for optimum localization. Our smoke detection results evince a noticeable gain up to 3% in accuracy and a decrease of 0.46% in False Alarm Rate (FAR), while segmentation reports a significant increase of 2% and 1% in global accuracy and mean Intersection over Union (IoU) scores, respectively. This makes our method a best fit for smoke detection and segmentation in real-world surveillance settings.
... A variety of visual smoke recognition and detection methods have been proposed in the last decades. Among these methods, different smoke features are adopted to determine the existence of fire, such as smoke color [1][2][3], motion [2][3][4][5], shape [6] and texture [7][8][9][10][11][12][13]. There are also a number of methods based on transform domains, and most of these methods use the information from wavelet transform to detect fire [14][15][16][17][18][19]. ...
It is a challenging task to accurately recognize smoke from visual scenes due to large variations in smoke shape, color and texture. To improve recognition accuracy, we propose a framework mainly with a robust local feature extraction module based on Gabor convolutional networks. We first propose a Gabor convolutional network, each layer of which mainly consists of Gabor convolution and feature fusion. To fuse feature maps generated by Gabor convolution, we present three Basic Grouping Methods, which divide the feature maps into several groups along orientation axis, scale axis and both of them. To avoid exponential growth of feature maps and preserve discriminative information simultaneously, we propose three element-wise aggregation functions, which are mean, min and max, to combine feature maps in each group. To further improve efficiency, we use local binary pattern to encode hidden and output maps of Gabor convolutional layers. In addition, we use a weight vector to optimize concatenation of histograms for further improvement. Experiments show that our method achieves very outstanding results on smoke, texture and material image datasets. Although the feature extraction step of our method is training-free, our framework consistently outperforms state-of-the-art methods on small smoke datasets, even including deep learning-based methods.
... Smoke detection methods are broadly classified into colorbased, motion-based, and hybrid methods [8]- [10]. For example, in [8], color information is combined with a motion for smoke detection. ...
... In [9], color features are combined with the image's energy in order to perform smoke detection. The work in [10] uses color information for smoke detection by employing fuzzy C-mean and back propagation neural networks. In addition to color and motion, other properties of smoke, such as its shape and other spectral, spatial, and temporal characteristics are investigated for its detection and classification, as given in [11]- [14]. ...
Smoke detection in foggy surveillance environments is a challenging task and plays a key role in disaster management for industrial systems. The current smoke detection methods are applicable to only normal surveillance videos, providing unsatisfactory results for video streams captured from foggy environments, due to challenges related to clutter and unclear contents. In this paper, an energy-friendly edge intelligence-assisted smoke detection method is proposed using deep convolutional neural networks (CNN) for foggy surveillance environments. Our method uses a light-weight architecture, considering all necessary requirements regarding accuracy, running time, and deployment feasibility for smoke detection in industrial setting, compared to other complex and computationally expensive architectures including AlexNet, GoogleNet, and VGG. Experiments are conducted on available benchmark smoke detection datasets, and the obtained results show better performance of the proposed method over state-of-the-art for early smoke detection in foggy surveillance.
... More specifically, Barmpoutis et al. [7] proposed a smoke detection method by converting the RGB space into the HSV colour space. Some other methods [8][9][10] converted the RGB space into YUV [8], YCbCr [9] or HSI [10] colour spaces. These methods are able to achieve some success when there are few interferences from other moving objects. ...
... More specifically, Barmpoutis et al. [7] proposed a smoke detection method by converting the RGB space into the HSV colour space. Some other methods [8][9][10] converted the RGB space into YUV [8], YCbCr [9] or HSI [10] colour spaces. These methods are able to achieve some success when there are few interferences from other moving objects. ...
Existing smoky vehicle detection methods and related smoke detection methods still have high false alarm rates. To solve this issue, this paper presents a contour-based smoky vehicle detection method for alarm systems. In this method, the hybrid background estimation method and some rules are adopted to detect vehicle objects. Three groups of artificial features based on the contour of the vehicle object are designed and extracted to distinguish between smoky vehicles and non-smoky vehicles. More specifically, by analysing the contour and convex hull at the back of the vehicle, some representative features, including the static features based on the contour and convex hull; statistical features of the mean, variance, skewness and kurtosis in the potential smoky region; and dynamic features of contour matching based on Hu moments, are designed and fused for smoky vehicle recognition. The support vector machine is adopted to classify the extracted features. The experimental results show that the proposed method can achieve better performance and lower false alarm rates than existing methods.
... The current literature shows that several researchers have presented smoke detection methods based on shape, color, texture and motion features. For instance, [15][16][17] used color based decision for detection of smoke regions. Chunyu et al. [15] proposed a smoke detection method based on motion and color features by making use of optical flow and back propagation neural network for classification of smoke. ...
... Calderara et al. [16] took advantage of image energy and color information for development of smoke detection system. Nguyen et al. [17] developed a smoke detection technique which uses fuzzy Cmean for clustring and back propegation neural network for the classificaiton of smoke based on color features. Besides color and motion, shape features are also exploited for smoke detection. ...
Smoke detection in IoT environment is a primary component of early disaster-related event detection in smart cities. Recently, several smoke and fire detection methods are presented with reasonable accuracy and running time for normal IoT environment. However, these methods are unable to detect smoke in foggy IoT environment, which is a challenging task. In this article, we propose an energy-efficient system based on deep convolutional neural networks (CNN) for early smoke detection in both normal and foggy IoT environments. Our method takes advantage of VGG-16 architecture, considering its sensible stability between the accuracy and time efficiency for smoke detection compared to the other computationally expensive networks such as GoogleNet and AlexNet. Experiments performed on benchmark smoke detection datasets and their results in terms of accuracy, false alarms rate and efficiency reveal the better performance of our technique compared to state-of-the-art and verifies its applicability in smart cities for early detection of smoke in normal and foggy IoT environments.
... Many researchers have investigated computer vision-based smoke detection technologies. Most of these are based on multistage pattern recognition, which basically consists of three stages [2][3][4][5]8,10,12,[14][15][16][17][18][19] : the preprocessing stage (PPS), feature extraction (FE), and classification (CLASSIFY). The PPS is the fundamental task used to determine the region of interest (ROI) in the input video frames by identifying smoke pixels and analyzing the eligible regions [8] . ...
... The PPS is the fundamental task used to determine the region of interest (ROI) in the input video frames by identifying smoke pixels and analyzing the eligible regions [8] . For segmenting the ROI from the input (derived from the static cameras), color segmentation (CS) [2,4,5,12,14,15,17,19] methods have been used by many researchers in various color spaces. This is commonly done by converting the RGB space into the HSV [2] , YUV [12], YCbCr [17] , or HSI [14] color spaces. ...
... For segmenting the ROI from the input (derived from the static cameras), color segmentation (CS) [2,4,5,12,14,15,17,19] methods have been used by many researchers in various color spaces. This is commonly done by converting the RGB space into the HSV [2] , YUV [12], YCbCr [17] , or HSI [14] color spaces. The HSV color space emphasizes visual perception in terms of variations in the hue, saturation, and intensity values of an image pixel, which makes segmentation easier to accomplish. ...
Detecting smoke during the initial stages is vital for preventing fire events. This study proposes a video-based approach for alarm systems that detects smoke based on temporal features extracted from optical smoke flow pattern analysis and spatial-temporal energy analysis. To do this, it considers various optical characteristics such as the diffusion, color, and semi-transparency of smoke. In the proposed model, smoke-colored pixels are identified via masking in the HSV color space and a temporal frame difference is applied. To extract the temporal feature vectors, we propose a new method that determines the optical flow of smoke by using distinguished texture information by applying a Gabor filter bank with preferred orientations. In addition, when applied to an image that has been temporal-differenced, the energy of the spatial frequencies is fed as another feature into the feature vector. Finally, these features are fed to a support vector machine (SVM) to discriminate our data more thoroughly and provide accurate detection of smoke. Experiments are carried out with benchmark datasets, which show that the proposed approach can work effectively without false alarms.
